Subject: Re: libkvm problems on alpha, but not sparc or i386 (netbsd-1-6)
To: Nathan J. Williams <nathanw@wasabisystems.com>
From: Greg A. Woods <woods@planix.com>
List: port-alpha
Date: 09/30/2004 13:45:04
[ On , September 30, 2004 at 01:30:09 (-0400), Nathan J. Williams wrote: ]
> Subject: Re: libkvm problems on alpha, but not sparc or i386 (netbsd-1-6)
>
> Do you have DEBUG or LOCKDEBUG or some similar non-default option in
> the kernel config file? Those can change the sizes of structures in
> ways that foil kvm-readers.

Damn!  I have both defined.  I should have remembered this (though in my
defence I'm going point out that most of the discussion about the issue
has apparently occured on tech-smp and I'm not subscribed to that list
yet :-).

Thanks very much for the clue!

LOCKDEBUG seems to be the magic one that causes the most "damage" to
kmem grovellers.  I don't think DEBUG even affects struct proc any more.
Unfortunately LOCKDEBUG adds a lot of size to the lock and simplelock
structures so I guess it's a bit too much to always include those
fields.....

Perhaps if we had a "kern.snafu" sysctl that would report the values of
DEBUG and LOCKDEBUG and such to libkvm then we could at least explain to
the user why these tools fail and/or give broken/meaningless results....

-- 
						Greg A. Woods
						Planix, Inc.

<woods@planix.com>     +1 416 489-5852 x122     http://www.planix.com/